TCP/IP versus Monte Carlo
How do you interpret a VC not getting back to you for a long time? There are two models from computer science that suggest two opposite things
When you are dealing with busy people such as venture capitalists, it is common that an email goes unanswered, or that someone who said they will follow up will fail to do so. While I fully empathise with that (having done that several times in the past), the issue with this is that the counterparty doesn’t know how to interpret this.
Thinking about it - there are two concepts from Computer Science that can help us think through this kind of a situation.
TCP/IP
Rather appropriately, from the world of computer networking comes TCP/IP (transmission control protocol / internet protocol, for the uninitiated). TCP/IP provides reliability of communication through what is known as the PAR (positive acknowledgement, resend) protocol.
The message is divided into packets (I wanted to insert a bite-sized pun here, but that may not be the most accurate), and sent one by one. When a receiver receives a package, it ought to acknowledge (“ACK”) the receipt of the package. The relevant part of PAR for us is that in case the sender fails to receive the ACK within a particular amount of time, the package is deemed to be “timed out” (not in the Angelo Mathews sense) . And then the package is resent.
This happens until there is a positive acknowledgement.
What does this mean for us, humans communicating in natural language with other busy humans? If the other person doesn’t respond within a certain period of time, follow up. And if the follow up isn’t responded to as well, after a few iterations, give up, assuming that the other person isn’t really interested in replying. In other words, if they haven’t replied to you in the time when they said they’ll reply to you (plus some buffer), it means your message has timed out, and they’re not interested. They are “ghosting” you.
Till recently, that is how I used to interpret delays in replying. This possibly comes from my experience in once unsuccessfully applying to, and then later working for, a company that is notorious for not sending formal rejections to applicants.
However, a couple of events / conversations in this week so far have made me reconsider this model. And instead switch to …
Monte Carlo Algorithms
Now we leave the world of computer networking and move to cryptography (in terms of my computer science undergrad at IIT Madras, I’m moving from 5th semester to 4th semester). One of the main issues in cryptography is primality testing - to check if a given large number is prime.
I won’t go into the details here, but this is harder than it sounds. And so computer scientists have come up with an ingenious way - given a number and a divisor, it is easy to verify that the divisor is a factor of the number. And if you are able to find one such divisor, then you know the number is not prime.
And so you guess possible divisors (using Fermat’s Little Theorem), and check in quick time if they divide the given “prime”. Each time you find a divisor that doesn’t divide the number, the probability of it being “prime” goes up. I’d written about this in a completely different context two decades ago.
The way computer scientists get around this peculiarity of this algorithm is by running it multiple times. They define a probability limit (say 95%) and say that “if I am 95% sure that it is a yes, then I’ll take it as a yes”. Now, the number of times the algorithm needs to be run in order to get 95% confidence is determined by the probability that the algorithm is right when it says “yes”.
Related to this, a venture capitalist I spoke to recently said, “if it is a no, we’ll quickly tell you that it is a no”.
Taking this statement at face value, one way to think of delays in VCs replying to you is in terms of a Monte Carlo algorithm. Let’s say they have said they will “do some research and get back to you”. There are some N people they need to talk to as part of their research, to figure out if your idea is spongeworthy.
The moment one of these N says “bad idea”, the VCs know you are not spongeworthy, and immediately send you a reject. However, they need to have their conversations with all of these N people (I have absolutely no idea what the distribution of N is, and what they talk about), and these people are also fairly busy people. So it takes time.
Under this model, until the VCs have spoken to all these N people and not got noes, they haven’t yet written you off as “unspongeworthy” (though the moment they get one no, you are out). And so the lack of response means that the conversation is still “alive”. Hopefully it will soon get to a stage where the VCs will “fail to reject” you, and you get some money, or another conversation!
Which of these frameworks do you think applies better here? How do you normally handle a lack of response in a conversation? How do you approach the tradeoff between following up too much and too little?
Does it matter for what you need to do? Either way, send a follow up mail after a polite interval. If they are interested but still not done with their N conversations, they won’t mind and will respond. If they are ghosting you, they won’t.
I guess the question is how much uncertainty can YOU handle. If, like in TCP/IP, you have to be certain, then send one or two follow ups. If you can handle the uncertainty, then Monte Carlo it is.